Archived version from NCDOCKS Institutional Repository http://libres.uncg.edu/ir/asu/

Diving Demand For Large Ship Artificial Reefs

By: Ashton Morgan, Matthew Massey, William L. Huth

AbstractUsing data drawn from a web-based travel cost survey, we jointly model revealed and stated preference trip count data in an attempt to estimate the recreational use value from diving the intentionally sunk USS Oriskany. Respondents were asked to report their: (i) actual trips from the previous year, (ii) anticipated trips in the next year, and (iii) anticipated trips next year assuming a second diveable vessel (a Spruance class destroyer) is sunk in the same vicinity. Results from several different model specifications indicate average per-person, per-trip use values range from $480 to $750. The “bundling” of a second vessel in the area of the Oriskany to create a multiple-ship artificial reef area adds between $220 and $1,160 per diver per year in value.

Ashton Morgan, Matthew Massey & William L. Huth (2009) "Diving Demand For Large Ship Artificial Reefs." Marine Resource Economics, 24(1), 43-59. Version of Record Available At www.bioone.org

Introduction On May 17, 2006, the ex-USS Oriskany, an Essex Class aircraft carrier was deliberately sunk off the coast of Pensacola, Florida to become “the world’s largest artificial reef.” The sinking was the culmination of two years of effort from a diverse set of individuals, institutions, and organizations. Its sinking created significant national media interest ranging from network coverage to a documentary film. It was hoped that the new artificial reef would provide many of the same ecosystem services supplied by a natural reef, including increased fish and sea-life habitat, improved fish stocks and angling quality, and new recreational diving opportunities (Adams, Lindberg, and Stevely 2006). If successful in providing these services, it is hoped the Oriskany will also relieve some of the use pressure on the area’s other reefs. Although the Oriskany’s effect on fish stocks and angling are unclear at this time, there have been thousands of divers who have visited the site in the year since its sinking.

The purpose of this paper is twofold. First, we estimate the non-market value of recreational diving on the Oriskany artificial reef, and second, we explore the potential value of adding additional artificial reefs to the area. To accomplish these tasks, we estimate several count data travel cost models based on combined revealed and stated preference diving trip counts to the Oriskany. As part of our modeling efforts, we also investigate the consistency of revealed and stated preference trip count data under varying site quality assumptions. The results provide the first estimate of divers’ willingness to pay for diving the Oriskany and should be transferable to other existing and potential large ship artificial reef sites. As the number of ships needing to be disposed of continues to increase, the value of creating artificial reefs and “bundling” additional vessels alongside existing artificial reefs to create multiple-ship reefs should also become increasingly important.

Data for the analysis are drawn from a web-based survey of individuals known to have dived the Oriskany in the year since its sinking. The survey asked respondents to report their: (i) actual Oriskany dive trips taken during the 2006 dive season, (ii) expected 2007 dive season trips under 2006 conditions, and (iii) expected 2007 trips assuming a second diveable warship is sunk in the vicinity of the Oriskany. Controlling for sampling method and diver characteristics, we combine the collected revealed and stated preference data and jointly estimate the relationship between trips demanded and travel cost and diver characteristics.

This paper proceeds as follows. First, we describe previous efforts to value recreational diving on artificial reefs. We then provide some background on the Oriskany and its sinking and describe our survey design and modeling strategy. Next we summarize our estimation data and present our results. We end with conclusions and recommendations for future work.

The Value of Recreational Diving Despite the recent significant growth in the number and popularity of artificial reef dive sites, there have been relatively few studies that focus specifically on artificial reef recreational diving use values. Broadening the scope to encompass studies including

any type of recreational diving valuation estimates increases the sample size, although a large percentage of the estimates are from studies that group values from multiple activities (fishing, diving, boating, etc.) or multiple dive site types (natural and artificial reefs). In many cases, the multiple-activity or multiple-dives site type estimates are not decomposable into accurate measures of divers’ artificial reef valuations (Kildow 2006). Roughly half of previous artificial reef valuation studies and reports of which we are aware focus on expenditure-driven economic impacts such as local output, employment, and labor income instead of on non-market recreational use values (Adams, Lindberg, and Stevely 2006). For valuation purposes, non-market estimates of dive site consumer surplus are theoretically preferred; however, the diving expenditure valuation literature does provide evidence suggesting the existence of substantial artificial reef recreational diving use values. For example, Bell, Bonn, and Leeworthy (1998) estimate the economic impacts from fishing and diving artificial reefs along the five-county region of northwest Florida to be approximately $461 million.3 Across the state in southeastern Florida, Johns et al. (2001) estimate that reef users spent approximately 10 million person-days using artificial reefs over a one-year period from 1997 to 1998, generating $2 billion in sales, $933 million in additional labor income, and 27,000 jobs in the region. Along the Texas coast, Ditton and Baker (1999) and Ditton et al. (2001) estimate that recreational expenditures of non-resident divers taking trips to the Flower Gardens Banks National Marine Sanctuary and other artificial reefs generated over $2.2 million in output at the local (coastal) level. Also in the Gulf of Mexico, Heitt and Milon (2002) estimate dives on oil and gas rigs result in total direct diving expenditures of $17.3 million and total economic activity of $32.5 million. Most recently and closely related to the Oriskany, Leeworthy, Maher, and Stone (2006) investigate the economic and ecological impacts of the 2002 sinking of the ex- USS Spiegel Grove off Key Largo in southern Florida. The authors estimate a net change in total recreational expenditures from pre- to post-deployment of $3.1 million. These new expenditures are further found to generate an additional $3.2 million in total output, $1.1 million in local income, and 68 new jobs. The majority of the non-market valuation consumer surplus estimates found in the diving literature use contingent valuation methods to elicit divers’ willingness to pay (WTP) for recreational diving opportunities, although several studies do employ travel cost models (Pendelton 2004). Both types of analysis may be seen in the handful of studies focusing explicitly on the recreational benefits of petroleum platforms. One of the first employs an iterative bidding process to estimate a mean WTP of $305 for an annual pass to dive petroleum rigs in the Gulf of Mexico (Roberts, Thompson, and Pawlyk 1985). Assuming an estimated diver population of 3,200, this implies a total annual use value of $976,000 for diving the rigs. Similarly, Ditton, and Baker (1999) and Ditton et al. (2001) test open- and closed-ended contingent valuation questions to estimate WTP for recreational reef diving off the coast of Texas. Their estimates range from $383 to $646 per year depending on the disclosure mechanism, with the closed-ended questioning providing larger estimates. In another example, McGinnis, Fernandez, and Pomeroy (2001) use a travel cost model to estimate the value of recreational diving and fishing platform Grace, an oil rig off the southern California

coast. They find a value of $68 per person per trip. With an average of three trips per year, the annual use value is $205 per person. We are aware of only three studies that have focused specifically on artificial reefs. Milon (1989) and Johns et al. (2001) both use contingent valuation questions to elicit use value for creating new artificial reefs. Milon estimates WTP for a new marine artificial reef site using several alternative incentive mechanisms and finds annual use values that range from $27 to $142. Johns et al. also utilize a contingent valuation methodology to estimate reef users’ value for maintaining artificial reefs in their existing condition and for investing and maintaining “new” artificial reefs. In the survey, respondents were informed of a proposed new artificial reef program with no specific mention of the vessels/ infrastructure that constituted the new reef. Results indicate diminishing marginal returns to increasing the size of the artificial reef system with annual use values per person for maintaining the existing reef of $75, compared to $24 for creating new artificial reefs. Finally, using dichotomous choice question responses from a sample of local and non-local users, Bell, Bonn, and Leeworthy (2006) estimate a total annual use value (not diving specific) of $25.0 million for artificial reef use across the Florida Panhandle region. The Oriskany Case Study The national defense reserve fleet was established after World War II to serve as an inventory of vessels available for use in national emergencies and for national defense. At the end of 2005 there were approximately 255 vessels in the fleet. Vessels are periodically examined and reclassified. During that process some are moved into a “non-retention” status and targeted for disposal. According to the U.S. Department of Transportation Maritime Administration (MARAD 2007) vessel disposal program report, there were well over 100 obsolete vessels scheduled for future disposal. Over the period from 2001 through 2006 some 72 ships, including the Oriskany and several other warships, were disposed of. There are a number of options available for ship disposal including vessel donation and sale, dismantling (domestic and foreign recycling/scrapping), sinking as an artificial reef, and deep-sinking in the U.S. Navy SINKEX Program.4 Hess et al. (2001) examined the disposal options for the fleet of decommissioned vessels that were stored at various naval yards throughout the country at the time and concluded that reefing was the best option available. In particular, Hess et al. note that if one focuses on the costs and offsetting revenues associated with domestic recycling, international recycling, and reefing disposal options, reefing is “very promising” and one of the “least expensive” disposal options available to MARAD and the Navy. Hynes, Peters, and Rushworth (2004) reiterated the potential benefits from the reef disposal option and suggested that communities might be willing to cost share in the disposal process due to fiscal benefits from use after reef establishment. The Oriskany was actually sunk in May of 2006 and commercial dive charters to the new reef began two days after the sinking.5 The ship is now located 22 nautical miles (a nautical mile covers 1.151 statute miles) south of Pensacola and operators along a 60-

mile stretch of the Florida Panhandle from Destin, FL, to Gulf Shores, AL, offer trips to the Oriskany (see figure 1).

Figure 1. Oriskany Permit Area Source: Florida Fish and Wildlife Conservation Commission, Tallahassee, FL. Most charter boat operators in the area run vessels that can take up to six divers, and a few run larger vessels capable of taking 16-20 divers at a time. There are also many private vessels that visit the reef for diving purposes. Given seas running up to about three feet, approximate travel times for a vessel out to the reef are between 90 to 150 minutes (a mean of about 2 hours). The Oriskany is sitting upright in about 215 feet of water and the bow of the vessel points due south. The flight deck is about 135 feet deep and the top of the island superstructure is about 70 feet deep. Most divers that visit the Oriskany are recreational divers that stay within 130 feet of the surface within no decompression limits. Recreational divers usually stay in the vicinity of the vessel’s island superstructure and make two dives on the Oriskany on a single trip. There is a large contingent of technical divers that visit the ship as well. These divers use dive profiles that involve greater depths, decompression, the breathing of various gas mixtures, specialized equipment, and penetration of the below flight deck interior. Technical diving is much more training and equipment-intensive than recreational diving, and all technical divers have a number of different advanced diving certifications. The ordinary recreational diver will usually have what is termed a basic or advanced open water certification, and some might be certified to dive simple nitrox gas mixes. Most operators require or recommend that the diver have at least the basic open water

certification and a minimum number of dives before performing an advanced dive like the Oriskany. Survey Design Because no formal records are kept on the total number of private and commercial dive trips taken to the Oriskany, the only plausible method available to value the recreational opportunity is to survey a known sample of the divers about their past and expected future trips. To define our sample of Oriskany divers, we obtained diver liability release forms from one of the most active dive shops that charters trips to the Oriskany. From the forms, 248 diver email addresses were identified. Each diver was sent an email describing the purpose of the study, the importance and confidentiality of all completed responses, and a link to a web-based survey instrument (see Little et al. 2006 and Champ 2003 for detailed discussions of web-based surveys). As an incentive to increase response rates, survey recipients were informed that participants would be entered into a random drawing in which three individuals would win a $150 gift certificate to cover the charter boat fee for their next dive. Five days after sending the original email, individuals that had not yet done so were sent a reminder to complete the survey. In total, we received 177 responses (a 71% response rate). As the focus of this research is on day trips, 43 individuals that only took overnight trips were not included in the final data set. Seventeen respondents that did not complete all the questions in the survey were also excluded, leaving 127 complete and usable responses (a 51% response rate). Along with some basic demographic and diver experience questions, the survey asked respondents three trip-count related questions–one revealed preference and two stated preference questions. The initial revealed preference (TRP_RP) question asked respondents to report the actual number of single day dive trips taken to the Oriskany in the year since its sinking. Following the question on past trips, individuals were asked to provide their expected number of trips to dive the Oriskany in the upcoming 2007 dive season (TRP_SP). Finally, respondents were presented with a description of a potential ship/artificial reef bundling scenario. Respondents were told that the U.S. Maritime Administration has a number of out-of-service military ships of various types that are being considered for use as artificial reefs in a variety of locations in U.S. coastal waters and that one possible scenario for reefing the ships was to create a “multiple-ship reefing area” by sinking a Spruance class destroyer in the permit area with the Oriskany. Respondents were provided with the destroyer’s dimensions and proposed sinking depth and could choose to view a map of the proposed sinking location if they desired. They were further informed that charter boats would pass close by the destroyer on their way out to, and back from, the Oriskany. This would create the option to dive the Oriskany on the first dive, and then during the surface interval travel to the new destroyer and dive it before returning to port. Respondents were asked: “If the new destroyer was sunk and available to dive today, do you think it would change the number of diving trips you expect to take to the Oriskany site (now including the additional destroyer) in 2007?” If respondents select “yes,” they were prompted to select how many more or less trips they would take in 2007. This selection was then used to

define the number of trips they would expect to take in 2007 given the presence of the second bundled destroyer (TRP_SP_DESTR). Tables 1 and 2 provide the definitions and descriptive statistics for the variables collected in the survey and used in the analysis. Several trip count characteristics immediately stand out. First, the average number of trips divers are expecting to take in the upcoming dive season (TRP_SP) exceed the average number of trips taken in the previous year (TRP_RP). Second, the expected number of dive trips nearly doubles with the addition of the destroyer (TRP_SP_DESTR), from slightly over two trips to almost four trips per year. The increases in both expected trip counts suggests an increase in demand for dive trips in the upcoming season. The sheer size of the Oriskany dive site, especially with the addition of a second ship, may lead divers to feel that multiple trips may be necessary to fully explore the vessels. However, portions of the expected trip increases may also be due to either hypothetical bias in the survey’s stated preference responses or habit formation among divers. Within the stated preference literature, hypothetical bias is a widely recognized issue. For example, two summary meta-analyses studies focusing on hypothetical bias by List and Gallet (2001) and Murphy et al. (2005) found evidence that values for non-market goods derived from stated preference survey techniques often significantly exceed the values derived from revealed preference methods. Similarly, a number of recreation demand studies focusing on recreator experience and habit formation have shown that past visits or experience have a positive effect on the probability of choosing to visit a site again in future choice occasions (Adamowicz 1994; Provencher and Bishop 1997; Moeltner and Englin 2004). Because the sample used in this study consists of divers who have previously dived the Oriskany, they may be more likely than the general diving public to dive the Oriskany in a future season. The travel cost data show that, on average, divers incur approximately $531 in costs per trip to dive the Oriskany.8 These costs may seem high, but they include significant diving-specific fees in the form of access and equipment rental or purchase. For example, the average charter boat fee to take a diver out to the Oriskany is reported to be $174 (including tip).9 Travel costs to the substitute site (TCSUB) are significantly higher, representing the lack of notable close substitutes to diving the Oriskany. Consideration of the socio-demographic data indicates that the average diver in the sample is 43 years of age, earns close to $100,000 per year in household income, and has over 11 years of diving experience. Finally, in our sample, 26% of respondents are technical divers (TECH_DIVE), with the remaining 74% considered recreational divers.

Table 1 Variable Definitions

Table 2 Descriptive Statistics

Estimation Methodology As is standard when valuing outdoor recreational trips at a specific definable site such as the Oriskany, this study relies on demand-based, single-site travel cost models. Travel cost models exploit the tradeoffs recreators make between site quality and visitation costs when choosing where, and how often, to recreate. In the model, the number of trips taken in the season is the quantity demanded. The travel cost for accessing the site is interpreted as the price (see Parsons (2003) for a detailed discussion of travel cost models). Because the dependent variable, actual/expected trips (y), is a nonnegative integer with a high frequency of small numbers, we rely on several count data model specifications in our attempt to estimate the travel cost relationship. Following Haab and McConnell (2003), the basic model may be written:

(1) where the number of trips taken by an individual in a season to the site, y, is assumed to be a function of a vector of personal and site characteristic explanatory variables, x. These explanatory variables include the travel cost to access the site, TCy; a vector of trip costs to potential substitute sites, TCSUB; individual’s income, INC; a vector of socio-demographic and dive experience variables, z, believed to influence the number of trips; and a site quality or site attribute measure, q. In this study, the z vector is assumed to include the AGE, YRS_DIVE, and TECH_DIVE variables, and the site quality or attribute measure, q, is assumed to include the DESTR variable. The y vector is constructed by pooling the three trip count measures (TRP_RP, TRP_ SP, and TRP_SP_DESTR). The joint estimation of revealed and stated preferences has the advantage of allowing the estimation of preferences for situations outside of historical experience, while anchoring the stated preference responses to actual behavior. The presence of the stated preference elicitation dummy, SP, should account for and measure any hypothetical bias present in the stated preference trip counts (Egan and Herriges 2006; Whitehead 2005). Because we only survey past participants, our revealed choice data are truncated at zero. We do not believe that endogenous stratification is an issue in our sample since the sample was derived from diver liability waivers collected over a full dive season. Unlike a typical onsite sampling strategy that collects information on only one (or a few) day(s) over the course of a season thereby likely under-sampling those individuals who visit infrequently, our sample is effectively collected on every day of the season and therefore correctly samples all avidity levels. The probability that an individual will take y trips is first assumed to take the truncation at zero corrected Poisson form:

(2) where the parameter λ is the expected number of trips and is assumed to be a function of the variables specified in the model. A detailed discussion of truncated count models may be found in Creel and Loomis (1990) and Haab and McConnell (2002). Usually, λ takes a log-linear form to ensure nonnegative trip counts and may be written:

(3) where the β’s are the coefficients to be estimated. To simplify estimation, we assume that respondents are using temporally constant preference parameters and decision criteria when making trip choices and that there is no correlation between individuals’ choices across the different count methods and scenarios.12 Combining equations (2) and (3) allows us to define the truncation corrected Poisson likelihood function:

(4) where n indexes individuals (n = 1 … N). This likelihood function is then maximized to recover estimates of the β parameters. Using the estimated coefficients, an average per-person, per-trip access value, or consumer surplus, for a trip to the site can be estimated. Consumer surplus, or CS, represents a measure of the value a diver places on diving the Oriskany and is the difference between total willingness to pay for the trips and total trip cost. From our log-linear model, consumer surplus can be calculated as:

(5) where TCy 0 is the individual’s trip cost, and TCy choke is the choke price that at which the number of trips declines to zero. Annual per-person consumer surplus values are calculated by multiplying the per-trip consumer surplus value by the average number of predicted trips per year, λ. It is also possible to calculate the change in consumer surplus due to a change in site quality (i.e., the addition of a destroyer to the site). For example, the annual marginal value of a change in site quality may be found by:

(6)

where λ * is expected trips with the quality change, and the estimated travel cost parameter associated with the new quality conditions. One potentially undesirable characteristic of the Poisson model is its restriction that the conditional mean and variance of the dependent variable, λ, are equal. In a recreation demand framework, this can be a limiting assumption as data on trips taken commonly exhibit overdispersion (i.e., the variance in trips is often greater than the mean). Ignoring overdispersion in estimation can lead to inefficiency due to the underestimation of standard errors. When dealing with truncated at zero data, the truncated Poisson model’s assumptions may be even more troublesome because the truncated model’s conditional mean is actually larger than the conditional variance. Therefore, if the underlying distribution is incorrectly assumed to be truncated Poisson, it can lead to both inefficient and inconsistent parameter estimates (Cameron and Trivedi 1998). When faced with overdispersion, the negative binomial model is a natural alternative since it allows for differences in the mean and variance and tests for overdispersion. The truncated at zero negative binomial model probability function, which results from a gamma distributed error term in the mean for an individual, can be expressed:

(7) where Γ denotes a gamma distribution and α is the overdispersion parameter. As with the Poisson model, equations (3) and (7) may be combined to specify a likelihood function which is then maximized to recover parameter estimates. Consumer surplus is computed analogously to the Poisson model. Estimation Results Columns one and two of table 3 provide the truncated at zero stacked Poisson and negative binomial models estimation results for our Model 1 specification (equation 3). Model 1 is our most basic and restrictive model specification because it assumes that the pooled data can be described by a single set of parameters. While estimates from the Poisson and negative binomial models are very similar, the negative binomials model’s positive and significant alpha value indicates that there is overdispersion present in the data. This overdispersion means that the Poisson model is misspecified, and the negative binomial model is the more appropriate of the two. Estimation results are presented for both models to illustrate their similarity, but all results are discussed in terms of the negative binomial model in the following sections.

Table 3 Truncation at Zero Corrected Poisson and Negative Binomial Models

As expected, TCy is negative indicating that divers living farther from the site and facing higher travel costs take fewer visits. The size of TCy implies that every dollar increase in the price of the trip to dive the Oriskany leads to a 1% decrease in expected trips. The positive, but insignificant, coefficient of the substitute site travel cost parameter, TCSUB, signals that Key Largo is at best a weak substitute for the Oriskany artificial reef. A lack of good substitutes might be expected given the Oriskany’s status as the world’s largest artificial reef. Turning to the diver-related characteristics, TECH_DIVE is positive and significant indicating that technical divers take more Oriskany dive trips than recreational divers. This makes sense for two reasons. First, the Oriskany is probably a more attractive dive to technical divers as they can reach the large flight deck level and below flight deck interior providing more opportunities for exploration. Second, all else equal, technical divers also take more aggregate dives per year in order to gain and maintain a “technical” rating. Results also suggest that trips increase with YRS_DIVE and INC,

although the relationships are not statistically significant. AGE is significant and negatively correlated with the number of trips, signaling that older divers take fewer trips. The coefficients on the variables controlling for elicitation method and quality changes are also positive and highly significant. The coefficient on SP indicates expected trip totals for the upcoming season collected through stated preference questions tend to be larger than past year revealed trip totals. The size of the increase in expected trips suggests that it is likely due to hypothetical response bias often prevalent in the stated preference methodology, although diver habit formation created by previous dives on the Oriskany could also be an influence. The positive DESTR coefficient indicates that diver preferences are sensitive to the scope of the dive sites and that adding a destroyer in the vicinity of the Oriskany would cause an increase in the number of expected trips. As pointed out by Boyle (2003), scope is generally not a problem in use value estimates such as recreation demand. Because welfare estimates are directly related to a model’s estimated travel cost coefficient, and previous studies have found that assuming a single preference structure when combining revealed and stated preference data embodying large changes in site attributes and quality can lead to biased estimates (Huang, Haab, and Whitehead 1997), we also estimate two additional negative binomial model specifications that allow travel cost preferences to vary across the different trip counts. The first additional model (Model 2) tests whether travel cost preferences vary across the revealed (RP) and stated (SP) preference counts by including a term interacting travel costs and the stated preference dummy variable. The model is formally written:

(8) Where is the marginal effect of the stated preference elicitation method on baseline (revealed) travel cost preferences. Similarly, we also test whether travel cost preferences change when a destroyer is added to the Oriskany dive site (Model 3). The model is given by:

(9) where is the marginal effect on baseline (no additional destroyer) preferences due to the addition of a destroyer.

Results for the varying travel cost models are presented in the last two columns of table 3. Two main results stand out. First, the travel cost and stated preference interaction is not significant in Model 2, suggesting that respondents are using the same travel cost preferences when evaluating revealed and stated preference trips. The SP dummy variable does, however, remain positive and significant in all models, implying that stated preference elicitation has a positive effect on total trips taken. Second, the travel cost and additional destroyer interaction is positive and significant in Model 3, signaling that when a major change in the scope or quality of a site occurs, such as the addition of a destroyer, it can affect the magnitude of the travel cost preference parameters recreators use to determine their expected number of trips. In this case, the addition of the second destroyer makes the Oriskany dive site more attractive and lessens the disutility associated with travel to reach it. Log likelihood ratio tests confirm that Model 3 is preferred to Model 1 with at least 97.5% certainty and preferred to Model 2 with at least 95% certainty. Lastly, we turn out attention to the consumer surplus estimates. Using the estimated parameters from Models 1 through 3, we first calculate the average predicted trip totals for existing baseline conditions (without the additional destroyer) and potential improved conditions (with the additional destroyer). For each scenario, estimates corrected for potential hypothetical stated preference bias are also derived (i.e., with SP=0). Next, we use the estimated travel cost parameters and the average predicted trip total to calculate the per-person, per-trip, and per-person annual consumer surplus values associated with existing baseline conditions at the Oriskany. Finally, the marginal per-trip and annual consumer surplus gains from sinking an additional destroyer are also calculated. All mean welfare estimates are presented with 95% confidence intervals constructed using the Krinsky and Robb procedure (Creel and Loomis 1991). Results are presented in table 4. Depending on the assumed structure of travel cost preferences and whether hypothetical bias is corrected for, per-person, per-trip consumer surplus values under baseline conditions range from $480 to $750, and annual values range from $305 to $866. As illustrated in equation 5, per-trip consumer surplus estimates are driven by variations in the estimated travel cost parameters across the models. For example, explicitly modeling the effect of an additional destroyer on estimated travel cost preferences through the TCDESTR variable in Model 3 leads to the lowest baseline consumer surplus estimates of any model. In fact, the mean per-trip consumer surplus estimates from Models 1 and 2 fall outside of Model 3’s 95% confidence interval in every case except Model 2’s hypothetical bias-corrected estimate. Although the variable capturing the effect, TCSP, is insignificant, Model 2 is the only model specification in which the estimated per-trip consumer surplus estimates are affected by correcting for hypothetical bias. The main effect of correcting for hypothetical bias is an approximate 50% decrease in the number of predicted trips. This expected trip decreases leads directly to lower annual consumer surplus estimates. The reduction in the number of expected trips to

less than one per year actually results in annual consumer surplus estimates that are lower than the per-trip estimates. Results further suggest that the addition of a destroyer to the area will lead to large gains in consumer surplus. In terms of the per-trip marginal value of an additional destroyer, Table 4 Per-person Consumer Surplus

Model 3 predicts that an additional destroyer would almost double the value of a trip. Only Model 3 is able to predict a change in per-trip consumer surplus values because it is the only model that allows travel cost preferences to vary between baseline and “with destroyer” conditions. All three models predict significant increases in the numbers of expected trips taken with the additional destroyer, which translate into large annual marginal consumer surplus values from the addition of the destroyer. In Models 1 and 2, the annual marginal consumer surplus values are roughly 60% of the baseline annual consumer surplus values in the uncorrected cases and 48% of the baseline annual values in the corrected cases. Because the annual marginal consumer surplus value from the destroyer in Model 3 embodies an increase in value per trip and an increase in the number of expected trips, it is two to three and a half times the size of the comparable estimates from Models 1 and 2 and nearly twice the size of the predicted annual consumer surplus under existing baseline conditions. To come up with a rough estimate of the aggregate Pensacola area diver consumer surplus, we use the 4,029 reported total diver trips chartered by all dive shops in the area in the year since its sinking as a conservative estimate of the diver population. Multiplying our estimated baseline annual per-diver consumer surplus estimates by our assumed diver population gives us a range of annual consumer surplus values from $1.2 to $3.5 million. The addition of a destroyer adds a marginal value between $900,000 and $4.7 million, indicating there is a significant economic value in bundling vessels to create large ship reefing areas. It is important to note that the 4,029 trip total does not account for trips made in private boats or trips made from other ports, which means our estimate almost certainly underestimates the true total annual consumer surplus. Although not directly comparable to other existing use value estimates because different reef systems are being valued, it is interesting to note that the estimates of this study are of roughly the same magnitude as a number of other estimates. For example, Johns (2004) estimates the annual value of $3.6 million associated with existing artificial reef use in Martin County, FL, and Bell, Bonn, and Leeworthy’s (1998) results indicate a total annual value of $2.2 million for artificial reef use across the Florida Panhandle region. In term of adding additional reefs, Johns et al. (2001) estimate a total willingness to pay of $4 million in southeast Florida. Conclusions This paper employs a web-based travel cost survey of divers to provide the first estimate of the diving demand for the ex-USS Oriskany. Respondents were asked to report both actual trips taken to the Oriskany in the year since its sinking and anticipated trips in the following dive season both with and without with the addition of a Spruance class destroyer to create a multiple-ship artificial reef. We jointly model stated and revealed preference trip count data using Poisson and negative binomial models controlling for sampling method and diver characteristics. The study finds that in this case revealed and stated preferences are suitable for combination, although care must be taken before assuming that stated and revealed

preferences can be described by a single set of parameters. We find consistent evidence of a significant hypothetical bias effect through a stated preference dummy, but also find that travel cost preferences do not vary significantly between stated and revealed counts. Large site quality changes, such as the addition of a destroyer to the dive area, are found to alter the preferences used to evaluate trip choices. Results also indicate that there are significant welfare benefits to divers from Oriskany- specific dive trips. The addition of a second destroyer to create a multi-ship artificial reef is found to add a significant amount of value and improve the desirability of the site. As MARAD seeks to dispose of more decommissioned vessels from its large inventory, the results of this study suggest that reefing is a valuable alternative and that bundling ships could provide extra value and disposal opportunities.

References Adamowicz, W. 1994. Habit Formation and Variety Seeking in a Discrete Choice Model of Recreation Demand. Journal of Agricultural and Resource Economics 19(1):19-31. Adamowicz, W., J. Louviere, and M. Williams. 1994. Combining Revealed and Stated Preference Methods for Valuing Environmental Amenities. Journal of Environmental Economics and Management (26):271-92. Adams, C., B. Lindberg, and J. Stevely. 2006. The Economic Benefits Associated with Florida’s Artificial Reefs. IFAS/EDIS Report, University of Florida, Gainsville, FL. Bell, F.W., M.A. Bonn, and V.R. Leeworthy. 1998. Economic Impact and Importance of Artificial Reefs in Northwest Florida. Office of Fisheries Management and Assistance Service, Florida Department of Environmental Administration. Boyle, K. 2003. Contingent Valuation in Practice. A Primer on Nonmarket Valuation, P.A. Champ, K. Boyle, and T. Brown, eds., pp. 111-70. AH Dordrecht, The Netherlands: Kluwer Academic Publishers. Cameron, A.C., and P.K. Trivedi. 1998. Regression Analysis of Count Data. New York, NY: Cambridge University Press. Champ, P.A. 2003. Collecting Survey Data for Nonmarket Valuation. A Primer on Nonmarket Valuation, P.A. Champ, K. Boyle, and T. Brown, eds., pp. 59-98. AH Dordrecht, The Netherlands: Kluwer Academic Publishers. Creel, M.D., and J.B. Loomis. 1990. Theoretical and Empirical Advantages of Truncated Count Data Estimators for Analysis of Deer Hunting in California. American Journal Agricultural Economics 72:434-41. _____. 1991. Confidence Intervals for Welfare Measures with Application to a Problem of Truncated Counts. The Review of Economics and Statistics 73(2):370-73. Ditton, R., and T.L. Baker. 1999. Demographics, Attitudes, Management Preferences, and Economic Impacts of Sport Divers Using Artificial Reefs in Offshore Texas Waters, Texas Parks and Wildlife Department, Austin, TX. Ditton, R., C. Thailing, R. Reichers, and H. Osburn. 2001. The Economic Impacts of Sport Divers Using Artificial Reefs in Texas Offshore Waters. Proceedings of the Annual Gulf and Caribbean Fisheries Institute (54):349-60. Egan, K., and J. Herriges. 2006. Multivariate Count Data Regression Models with Individual Data from an On-site Sample. Journal of Environmental Economics and Management (52):567-81. Grijalva, T.C., R.P. Berrens, A.K. Bohara, and W.D. Shaw. 2002. Testing the Validity of Contingent Behavior Trip Responses. American Journal of Agricultural Economics 84(2):401-14. Haab, T.C., and K.E. McConnell. 2002. Valuing Environmental and Natural Resources: The Econometrics of Non-market Valuation. Northampton, MA: Edward Elgar.

Heitt, R., and J.W. Milon. 2002. Economic Impacts of Recreational Fishing and Diving Associated with Offshore Oil and Gas Structures in the Gulf of Mexico. Department of the Interior Mineral Management Service, Washington, D.C. Hess, R., D. Rushworth, M. Hynes, and J. Peters. 2001. Disposal Options for Ships. Rand Monograph Report. Rand Distribution Services, Santa Monica, CA. Horn, B., J. Dodrill, and K. Mille. 2006. Dive Assessment of the Oriskany Artificial Reef. Division of Marine Fisheries Management Artificial Reef Program, FWC, Tallahassee, FL. Huang, J.-C., T.C. Haab, and J.C. Whitehead. 1997. Willingness to Pay for Quality Improvements: Should Revealed and Stated Preference Data Be Combined? Journal of Environmental Economics and Management (34):240-55. Hynes, M., J. Peters, and D. Rushworth. 2004. Artificial Reefs: A Disposal Option for Navy and MARAD Ships. RAND, National Defense Research Institute, Santa Monica, CA. Johns, G. 2004. Socioeconomic Study of Reefs in Martin County, Florida. Prepared for Martin County, FL, by Hazen and Sawyer, P.C., Hollywood, FL. Johns, G., V.R. Leeworthy, F.W. Bell, and M.A. Bonn. 2001. Socioeconomic Study of Reefs in Southwest Florida. Miami, FL. Prepared for Miami-Dade County, FL, by Hazen and Sawyer, P.C., Hollywood, FL. Kildow, J. 2006. Phase 1 Florida’s Ocean and Coastal Economies Report. National Oceans Economics Program, Moss Landing, CA. Leeworthy, V.R., T. Maher, and E. Stone. 2006. Can Artificial Reefs Alter User Pressure on Adjacent Natural Reefs? Bulletin of Marine Science 78(1):29-37. List, J.A., and C.A. Gallet. 2001. What Experimental Protocol Influence Disparities Between Actual and Hypothetical Stated Values? Environmental and Resource Economics 20:241-54. Little, J.M., K.M. Grimsrud, P.A. Champ, and R.P. Berrens. 2006. Investigation of Stated and Revealed Preferences for an Elk Hunting Raffle. Land Economics 82(4):623-40. McGinnis, M., L. Fernadez, and C. Pomeroy. 2001. The Politics, Economics, and Ecology of Decommissioning Offshore Oil and Gas Structures. Department of the Interior Mineral Management Services, Washington, D.C. Milon, J.W. 1989. Contingent Valuation Experiments for Strategic Behavior. Journal of Environmental Economics and Management 17:293-308. Moeltner, K., and J. Englin. 2004. Choice Behavior Under Time-Variant Quality: State Dependence Versus “Play-It-by-Ear” in Selecting Ski Resorts. Journal of Business and Economic Statistics 22(2):214-24. Murphy, J.J., P.G. Allen, T.H. Stevens, and D. Weatherhead. 2005. A Meta-Analysis of Hypothetical Bias in Stated Preference Valuation. Environmental and Resource Economics 30:313-25.

Parsons, G.R. 2003. The Travel Cost Model. A Primer on Nonmarket Valuation, P.A. Champ, K. Boyle and T. Brown, eds., pp. 269-39. AH Dordrecht, The Netherlands: Kluwer Academic Publishers. Pendelton, L. 2004. Creating Underwater Value: The Economic Value of Artificial Reefs for Recreational Diving. San Diego Oceans Foundation, San Diego, CA. Provencher, B., and R. Bishop. 1997. An Estimable Dynamic Model of Recreation Behavior with an Application to Great Lakes Angling. Journal of Environmental Economics and Management 33(1):107-27. Roberts, J., M.E. Thompson, and P.W. Pawlyk. 1985. Contingent Valuation of Recreational Diving at Petroleum Rigs, Gulf of Mexico. Transactions of the American Fisheries Society 114(2):214-19. U.S. Department of Transportation Maritime Administration (MARAD). 2007. Report to Congress on the Progress of the Vessel Disposal Program, Washington, D.C. Whitehead, J.C. 2005. Environmental Risk and Averting Behavior: Predictive Validity of Jointly Estimated Revealed and Stated Behavior Data. Environmental and Resource Economics 32:301-16.